While there are many effective options for treating a major depressive episode, there are no clinical markers that predict the likelihood of remission with an initial trial of either an antidepressant medication or psychotherapy. More critically, there are also no reliable predictors that might anticipate failure to such standard treatments either alone or in combination. In prioritizing a role for direct measures of brain functioning in the development of new algorithms for clinical management of depressed patients, a systematic characterization of pretreatment patterns predictive of unambiguous remission and non-remission to standard treatments is a necessary first step. This project will characterize imaging-based brain subtypes that distinguish groups of depressed patients who later remit or not to SSRI pharmacotherapy or cognitive behavior therapy (CBT), respectively. To define these subtypes, a prospectively-treated cohort of 100 patients will be randomized to receive either escitalopram (s-CIT) or CBT for the first 12 weeks, with non- remitters to either first treatment crossed over to receive an additional 12 weeks of treatment with the alternative intervention. Non-remitters to both treatments will thus define a relatively treatment resistant third subgroup. Resting-state 18F-fluoro-deoxyglucose (FDG) positron emission tomography (PET) scans will be acquired prior to initiating antidepressant therapy, with pre-treatment scan patterns associated with three possible outcomes (CBT remission, s-CIT remission, and non-remission to both) assessed using multivariate analytic methods. A second PET scan, acquired early in the treatment course, will be used to assess the likelihood of response to the specific treatment first assigned. The proposed studies are a first step towards defining brain-based biomarkers predictive of differential treatment outcome in major depression;most critically, patterns distinguishing patients at risk for treatment resistance. Identification of such biomarkers has additional implications for future testing of novel therapies in patients with distinct brain signatures, including development of evidence-based treatment algorithms for individual patients.